Here's a paper that concludes TOR is hopeless against a network adversary given enough time, specifically: "Some of our results against an adversary controlling ASs or IXPs are similarly alarming. Some users experience over 95% chance of compromise within three months against a single AS or IXP." ... "An adversary with additional ASes or IXPs has much higher compromise speed, notably against even those users in “safer” locations. Such an adversary is highly relevant in today’s setting in which many large organizations control multiple ASes or IXPs." However, among the conclusions there is also this promising one: "Surprisingly, we observe that high diversity in destinations may actually result in improved security against a network adversary." May be re-inventing the wheel here, but what if lots of scripts pretending to be firefox are made to browse a high diversity of destinations, for example retrieved from search engine results with search terms picked randomly from an offline dictionary over and over? With some country-specific searches too to diversify geographically? In other words, as you browse, you also run a bot with extremely diverse interests in many country domains. How effective would this be?